Taxonomic status of Tamarinus imperator subgrisescens (Lönnberg, 1940) (Cebidae, Callitrichinae)

Abstract The emperor tamarin, Tamarinus imperator, is composed of two subspecies, the nominal type, T. i. imperator, distributed between the Acre and Purus Rivers, whose range is limited between the Brazilian state of Acre and Peru are unbounded, and T. i. subgrisescens, occurring in Peru, Bolivia, and Brazil, in the Brazilian states of Acre and Amazonas. Morphologically, both taxa are easily identifiable by the pelage pattern (chromogenetic fields), and even being easily distinguishable, both lineages are considered subspecies according to the criterion based on the Biological Concept of Species from the 1970s, even without presenting some necessary criteria, such as the intergradation zone. Here we analyzed pelage traits, cranial morphometry, Cytochrome-b divergence, and distributional pattern data applying the premises of integrative taxonomy to elucidate the taxonomic status of both lineages. We hypothesize that both lineages are considered full species through a series of criteria for species recognition, such as distinguishability, level of phenotypical divergences of several morphological complexes with congruence among them, and some genetic divergence. The hybridization is unknown and the low or the lack of sampling in target areas does not allow us to determine whether a hybridization or even contact zone between the two lineages exists indeed. All character sets analyzed were congruent with each other and reinforced the high level of divergences between the two subspecies including several pelage differences, morphometry (descriptive statistics, PCA, and MANOVA), and mitochondrial DNA Cytochrome-b divergence. Most of the distribution in both lineages are allopatric, and the levels of intra-lineage phenotypical variation are much lower than between the lineages.


INTRODUCTION
Marmosets, Tamarins, and Goeldi's monkeys' treatise of Hershkovitz (1977) defined the taxonomic arrangement of the family Callitrichidae for decades, but within the last thirty years, the systematics of this small-bodied Neotropical primate group is still found in intense debate (e.g., Rylands et al., 2016) to elucidate phylogenetic relationships inside the genera and mainly to delimit the taxa at the species level.Many of the lineages are still considered subspecies, a conservative scenario from the 1970s based on the Biological Concept of Species (de Queiroz, 2007) and does not take into account criteria, such as the verification of intergradation zones.Non-hybridization zones can be due to several reasons, such as the absence of samples or studies in target areas or the real absence of hybridization.
In respect to Saguinus (lato sensu), Hershkovitz (1977) recognized 10 species, clustering them into three species groups based on the facial morphology: haired, mottled, and bare-faced tamarins.Based on the high level of molecular divergence, Cropp et al. (1999) suggested that Saguinus should be divided into two genera: Leontocebus, grouping the small-bodied species (= S. nigricollis group); and Saguinus, for the remaining species.This arrangement was formalized later by Rylands et al. (2016).Later, Garbino & Martins-Junior (2017) divided Saguinus into three species groups and ranked them at the subgeneric level based on morphological and molecular markers: Saguinus, Leontocebus, and Tamarinus.Some authors refused Leontocebus and Tamarinus as distinct genera; the subgenus Tamarinus as recognized included S. imperator, S. labiatus, and S. mystax.Lastly, Brcko et al. (2022) again revised the tamarins' supraspecific taxonomic arrangements and based on a phylogenetic analysis of 44 nuclear and mitochondrial markers, restructured the group elevating once again Leontocebus to the full genus rank and proposed three species groups for Saguinus, each one representing a different lineage that would represent a distinct genus: Tamarinus (including inustus and mystax groups), Saguinus (bicolor and midas groups), and Oedipomidas (oedipus group).
Tamarinus imperator (Goeldi), following the last generic arrangement proposed by Brcko et al. (2022), is one of the seven polytypic species for tamarins (Rylands et al., 2016).Its distribution ranges from southeastern Peru and northwestern Bolivia to western Brazil in the states of Amazonas and Acre (Rylands et al., 1993).Tamarinus imperator is defined by a conspicuous, long, white mustache, and two subspecies are recognized.Tamarinus i. imperator, occurs between the right bank of the Purus River and the left bank of the Acre River.This subspecies is represented by scarce records, and its original description was based on a low number of voucher specimens available both in collections and field observations.Tamarinus i. subgrisescens (Lönnberg, 1940) occurs from the right bank of the upper Juruá River, in the Brazilian states of Amazonas and Acre, to the upper Ucayali Basin, Peru (Hershkovitz, 1979), andPando, Bolivia (Buchanan-Smith et al., 2000); it is better represented by vouchers in collections, and there are more available data of field records and ecology.
Tamarinus i. imperator and T. i. subgrisescens were delimited by Hershkovitz (1977Hershkovitz ( , 1979) ) using discrete pelage traits and parapatric distribution.Both taxa are easily distinguishable by the color of dorsal and tail pelage, and the presence or not of a white beard.No hybridization zone was evident, one of the premises to support the subspecific level.Indeed, no clear barrier is observed separating both taxa by their distribution, and the limit of these subspecies along the upper Acre, Purus, and Iaco Rivers, in Acre, is unclear.
Here, we used a comparative analysis of pelage, skull morphometrics, DNA evidence, and the distributional pattern to clarify the taxonomic status of these two lineages.Even with the low sampling of T. i. imperator, our study brings novelties regarding new diagnostic pelage characteristics to define both taxa besides consistent Cyt-b divergence among three individuals representing both taxa and new evidence on the morphology of type series material of T. i. imperator.We also updated the distribution of both taxa based on additional studied specimens and other sources of information as personal and literature reports.

Material examined
The analyzed material included skins (denoted by sn) and skulls (denoted by sl) of 66 specimens housed in the following institutions: American Museum of Natural History, New York, USA (AMNH), Instituto Nacional de Pesquisas da Amazônia, Manaus, Brazil (INPA), Museu Nacional, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil (MNRJ), Museu Paraense Emílio Goeldi, Belém, Brazil (MPEG), Museu de Zoologia da Universidade de São Paulo, São Paulo, Brazil (MZUSP), Museum of Vertebrate Zoology, Berkeley, USA (MVZ), Swedish Museum of Natural History, Stockholm, Sweden (NRM), and The Field Museum, Chicago, USA (FMNH).We directly analyzed 60 specimens, including the lectotype of T. i. imperator housed at the MPEG, and five topotypes of T. i. subgrisescens housed in MZUSP and MNRJ.Six specimens, including all type series of T. i. subgrisescens (lectotype and paralectotypes), were studied using photographs sent by researchers and curators.

Pelage
We carried out the study of pelage coloration considering the chromogenetic fields as defined by Hershkovitz (1977).The chromogenetic fields were the head (crown), face, chin, scapular region, mid-dorsal portion (back), rump, chest, belly, inner and outer sides of the fore and hindlimbs, and the dorsal and ventral portions of the tail.Differential distribution of hairs on the chin and upper border of the ears was also analyzed.
Before performing the multivariate analysis, we evaluated the sexual dimorphism of each variable by applying a Shapiro-Wilk test (Table 2) to check the normality of each variable, and then we performed a t-test (parametric) or a Mann-Whitney U (non-parametric) according to the distribution of the variables.We consider p ≤ 0.05 as significant for all statistical tests.Because the three specimens of T. i. imperator are male, statistical tests verifying sexual differences were applied only to T. i. subgrisescens (11 females and nine males).Subsequently, we performed a Principal Component Analysis (PCA) to verify the distribution of the set of variables for each individual at the vectorial space using a variance-covariance matrix (Cadima & Jolliffe, 1996).Lastly, we applied MANOVA using Hotteling's value with Bonferroni corrected.Among the three skulls of T. i. imperator, one was partially damaged, thus for multivariate analyses, we worked with a reduced dataset Pap.Avulsos Zool., 2023;v.63: e202363005 3/14 composed of nine measurements (variables 4, 5, 6, 7, 8, 9, 11, 12, and 13) to include all three skulls of T. i. imperator for the multivariate analyses.Both analyses as described above were performed in PAST© version 4.0.

Mitochondrial DNA
We used two mitochondrial markers widely available in Genbank, Cytochrome-b and 16S, for estimative, molecular, phylogenetic relationships.We extracted DNA from the muscle of two T. i. imperator specimens (Table 1) through the phenol-chloroform method (Sambrook & Russel, 2001).We re-suspended the extracted DNA in 50 μL of TE buffer.We amplified the 5'region of the Cytochrome-b gene using the primers CytB1 AATGATATGAAAAACCATCGTTGTA and Cytochrome-b TTTCAGCTTTGGGTGTTGATG (Matauschek et al., 2011).That of the 16S was amplified using the primers L1987-5′ GCCTCGCCTGTTTACCAAAAAC 3′ and H2609-5′ CCGGTCTGAACTCA GATCACGT 3′ (Araripe et al., 2008).We amplified both genes in a 25 μL polymerase chain reaction (PCR) mix, including 0.3 units of Taq DNA polymerase (Invitrogen Platinum™ Taq DNA Polymerase), 2 mM MgCl₂ in 1× PCR buffer, 0.5 μMol per primer, 2.5 mM dNTPs, and about 20 ng of genomic DNA.We carried out the amplification in a thermocycler using a program consisting of 5 min of denaturation at 94℃, followed by 37 cycles -30 seconds at 94℃, 30 s at 50℃, 1 min at 72℃, and a final extension for 10 min at 72℃.We visualized the PCR products in a 2% agarose gel.Subsequently, we purified the products of positive reactions using polyethyleneglycol 20% (PEG 20%) (Santos-Júnior et al., 2015), we sequenced the purified PCR products in both directions with the same primers used in the PCR using the BigDye terminator sequencing kit (Applied Biosystems, Waltham, Massachusetts), and then we analyzed them with an ABI 3130xl (Applied Biosystems).We obtained the consensus sequences with SeqScape v.2.6.All laboratory procedures were conducted at the Laboratório de Biodiversidade e Evolução Molecular, Universidade Federal de Minas Gerais, Brazil (LBEM).In addition to the three sequences generated here (one CytB, and two 16S sequences), 40 Cytochrome-b and 16S sequences available in the GenBank were employed for 20 species belonging to Callitrichinae (Table 1).
For molecular data analyses, we aligned each gene individually using the algorithm MAFFT in the online server (https://mafft.cbrc.jp/alignment/server;Katoh et al., 2017) assuming the default parameters.Subsequently, we submitted the 16S sequences to Gblocks 0.91b (Castresana, 2000) to search for and remove regions with ambiguous alignment.We used the default settings, except for the "allowed gap positions" option, which was set to 'with half' .After that, we concatenated the "two genes" alignments using the software SequenceMatrix v.1.8(Vaidya et al., 2011).We explored the best partitioning schemes and substitution models simultaneously, using PartitionFinder v.2.1.1 (Lanfear et al., 2017) under a Bayesian information criterion for the entire matrix.The branch lengths were unlinked, the criterion for model selection was the corrected Akaike information criterion, and the search was done for all possible partitioning schemes.
We performed the phylogenetic analyses using a concatenated matrix with Cytochrome-b and 16S gene data, including a total of 20 species.We generated the phylogenies through Bayesian Inference (BI) in MrBayes 3.2.7 (Ronquist et al., 2012) using two sets of Markov chains, each containing three hot chains and one cold, temperature set to 0.05 with 20 million generations and a 25% burn-in, to seek for convergence to the same subset of best trees.Convergence of the runs was assessed using the following statistics: standard deviation of split frequencies, potential scale reduction factor (PSRF), and estimated sample size (ESS) for each parameter in Tracer 1.7.1 (Rambaut et al., 2018).We obtained the interspecific genetic distances with Mega X, using the parameters of the Kimura 2 model-K2P and variance estimation Bootstrap method with 500 replications (Kumar et al., 2018).We performed two analyses with different data sets, one using the Cytochrome-b gene (1010 pb) and the other using the 16S gene (459 pb).We chose to use a data set for each gene because when both were used in the same data set, the amount of missing data prevented the achievement of a satisfactory result.

Pelage coloration
We confirmed the five differences in the chromogenetic fields as already described by Hershkovitz (1979), in which the author delimited both subspecies.They are 1) a chin with just two patches of short, white hairs in T. i. imperator (Fig. 2) and a tufted chin with long, white hairs in T. i. subgrisescens (Fig. 2); 2) a chin markedly triangular and blackish in T. i. imperator (Fig. 2) and incipient in T. i. subgrisescens (Fig. 2); 3) brownish mingled with white hairs on both the chest and belly in T. i. subgrisescens (Fig. 3) and reddish, orange, and white hairs mixed in T. i. imperator (Fig. 4); 4) the back and lateral fringe of hairs are dark agouti and brownish or somewhat yellowish in T. i. subgrisescens, and light grayish or buffy fringe of hairs in T. i. imperator; and 5) silvery, light brown-grayish hairs along the inner side of forelimbs in T. i. subgrisescens (Fig. 2) and silvery orange in T. i. imperator (Fig. 2).
In addition, we observed three other consistent pelage traits that distinguish individuals of T. i. subgrisescens and T. i. imperator: 6) grayish or whitish agouti on the rump and base of the tail pelage, dorsally, similar to the thighs in T. i. imperator (Figs. 3 and 5) and orange in T. i. subgrisescens (Figs. 4 and 5); 7) tail, in general, predominantly grayish-brown or blackish with some yellowish or whitish hairs mixed (Figs. 3 and 4), in T. i. imperator, but orange or rufous on the ventral side close to the genitalia; rufous tail, orange on the tail with pelage from the middle and distal portions that eventually mingled with scattered blackish hairs, mainly at the terminal hairbrush, in T. i. subgrisescens (Figs. 4 and 5); and 8) the presence of a long (11-12 mm), dense tuff of black hairs on the upper and posterior border of the ears in T. i. imperator (Fig. 6) and naked or sparsely haired in T. i. subgrisescens (Fig. 6).

Skull morphometry
Skulls of S. i. imperator proved to be smaller than T. i. subgrisescens in all measured variables (Fig. 7, Table 3).Even considering the discrepancy in the sampling between the two subspecies, it is possible to notice that the level of divergence of the measures between T. i. imperator and T. i. subgrisescens is larger than other sister species pairs, such as S. niger and S. umbratus (Gregorin & Vivo, 2013) or L. cruzlimai and the subspecies of L. fuscicollis (L.f. avilapiresi, L. f. primitivus, L. f. mura, L. w. weddelli, and L. w. melanoleucus -Sampaio et al., 2015).The scores of the PCA and MANOVA are in Table 4. Figure 8 shows that in both analyses (PCA and MANOVA), there was the formation of two clusters on axis 1, representing S. i. imperator and S. i. subgrisescens.Axis 1 is influenced by size and shows the differences in dimensions between the two groups of individuals representing the two already recognized lineages, with S. i. imperator being the smaller form.The variables that most contributed to this grouping in PCA were the height and length of the mandible.Axis 2 indicates some level of variation in S. i. subgrisescens when comparing samples from the Brazilian Amazon and Peru.

Molecular analyses
We removed a total of 74 pb of ambiguous alignment from the 16S gene (533 pb with the ambiguous alignment regions).The concatenation of the genic regions resulted in a matrix with 1599 pb (Cytochrome-b -1140 pb and 16S -459 pb, Table 5).The best partitioning schemes and substitution models are presented in Table 6.
Even considering that the low sampling of T. imperator does not permit testing for monophyly, the most important information for this study is the levels of genetic divergence, particularly Cyt-b among the three individuals of T. imperator, each representing the subspecies.The smaller genetic distances among pairs of evolutionary closed taxa of tamarins using Cyt-b were T. labiatus × T. mystax (4.31) and T. i. imperator × T. i. subgrisescens (4.42) (Table 7).Cytochrome-b has been a very useful mitochondrial marker for defining species in mammals, though defining them based uniquely on only one gene is not easy (Vallinoto et al., 2006).Cropp et al. (1999), using a poll of three mitochondrial regions (D-loop, Cytochrome-b, and 16S) found some similar genetic divergence levels for some sister species as compared to our data, such as 3.4 for S. midas × S. niger, 5.1 for L. tripartitus × L. nigricollis, and 4.9 for O. oedipus × O. geoffroyi.Thus, strictly regarding Cyt-b, our data is following the literature, and they are compatible in recognizing both taxa of T. imperator as full species based on genetic divergence.However, we reiterate that the use of a few individuals of T. imperator for molecular analysis is insufficient to express the intragroup variation and the possibility of these divergences being altered.On the other hand, this study is the first to provide comparative sequences between the two subspecies of T. imperator and in congruence with the other characters, the genetic data becomes informative.
Table 5. Full-length concatenated alignment data was used in the phylogenetic analysis.Numbers between parenthesis followed by 'N' refer to nucleotide base is not known.Applying the unified concept of species (de Queiroz, 2007) with several sources of evidence not only provides robustness to the delimitation of lineages, and therefore species, but also allows inferences about more complex speciation processes to observe, such as parapatric speciation (Gao et al., 2019).Considering that parapatric speciation occurs usually by the environmental gradient (dispersion - Florio et al., 2012) and not by geographic/environmental ruptures as commonly postulated for allopatric speciation (Gao et al., 2019), ecological studies of T. imperator and T. subgrisescens might corroborate our hypothesis of two lineages evolving independently (Blanckaert et al., 2020).The second challenge is the delimitation of the contact zone between the two taxa on the Acre-Peru border and checking for the presence or absence of the hybridization zone, which could be due to secondary contact or evolutionary processes that could be recent and ongoing.

Type locality and type series
Rio Acre, opposite Bom Lugar, state of Amazonas, Brazil.In the description of this species, Goeldi (1907) mentioned two juvenile specimens from Rio Acre and three (two adults and one infant) from upper Purús River, and no type specimen was designated.The original label of the female lectotype indicates "upper Rio Purus, Brazilian State of Amazonas, in Bom Lugar or perhaps Monte Verde".Several subsequent attempts to restrict the type locality for T. imperator resulted in quite confusing scenarios and were summarized by Hershkovitz (1979) and Rylands et al. (1993).Elliot (1913) mentioned "Rio Purus, a tributary of the Amazon, western Brazil", and Lönnberg (1940) restricted it to "the upper Rio Purús".Cabrera (1958) considered the type locality of T. imperator as "Río Acre y Purús; aqui restringida ao río Acre", but criticized by Carvalho (1959), who stated that "upper Rio Purus" would be more realistic.Nonetheless, the confusion persists due to the uncertainty of the provenience of the lectotype that could be placed on either locality as indicated in the original label of the lectotype, the Purus (Monte Verde) and Acre (Bom Lugar) Rivers.
The restriction made by Carvalho (1959) was cautiously reviewed by Hershkovitz (1979) once Bom Lugar and Monte Verde would be out of the distribution range of T. imperator: Bom Lugar is placed at the right bank of the Acre River whereas Monte Verde at the western (left) margin of the Purus River.However, Hershkovitz (1979) stated that specimens could be collected on the opposite bank of the Acre and Purus Rivers, and those towns were used as the nearest reference.The same procedure was done regarding the records of T. imperator from Manoel Urbano, left margin of Purus River, State of Acre, collected by P.E.Vanzolini in September 1973 (MZUSP 11238 and 11239).We checked the travel diary of Vanzolini who collected material on both banks along the Purus River and there is no indication on which side of the Purus River the individuals of T. imperator were collected, except he indicated the town of Manoel Urbano as a sampling reference.Recently, Rylands et al. (2016) reinforced that neither Monte Verde nor Bom Lugar should be the type locality of the black-chinned emperor tamarin.
The confusion of type locality is because the material housed at MPEG, one adult female (MPEG 914), one adult male (MPEG 36604), and one infant (MPEG 868) of T. imperator, which Goeldi (1907) described, was collected by Snethlage (1908) in a long expedition through the upper Purus region, including collection from both Monte Verde (Purus River) and Bom Lugar (Acre River), places relatively close to Amazon dimensions but that may have relevant biogeographical significance depending on the side of the river to be considered.Two juveniles of the five specimens that Goeldi (1907) had in his hands were sent to European museums and three were housed at the MPEG as stated by Goeldi (1907: 94) "the family with three individuals (♂♀ adults, and a young one) are already mounted in the Pará Museum".The label of one adult female (lectotype) stated the provenience as "Alto Rio Purús", in Monte Verde or Bom Lugar.After Goeldi, subsequent researchers that studied the type series from MPEG referred only to the female and the infant, but never mentioned the adult male (Carvalho, 1959;Hershkovitz, 1979).A detailed analysis of the adult male studied by us and housed in the type cabinet at MPEG revealed that the specimen presented a phenotype typical of T. i. subgrisescens, with a long, white mustache, brown chin, and a completely rufous tail.We observed that those traits have been previously noted by Goeldi (1907: 94) in his fig.23, which illustrated a male individual with two long tufts of hairs on each side of the face (a "mustache" and a "beard") and stated: "white hairs of the circumbuccal zone extending over the whole area of the lower jaw, not including the chin and inferior side of the jaw.As a result, the old male appears bearded as well as mustached" [Goeldi (1907: 95)], and "Most aberrant is the coloring of the tail in the old male.From the very insertion, the bright rusty-red color predominates in its whole circumference throughout its entire length except for a dark terminal tuft" [Goeldi (1907: 96)].The completely red tail of the adult male is shown in Goeldi's colored figure that was also reproduced by Sampaio et al. (2015).
Thus, we are hypothesizing that the three specimens that Goeldi (1907) had in his hands did not form a family as he thought, and there is a possibility that the adult male and female came from distinct places, perhaps the female from the Acre River (typical T. imperator) and the male with the phenotype of T. subgrisescens may be from the Purus River (there is no information on the label of specimen MPEG 36604 and the original one was probably lost).Therefore, the type locality of T. imperator could be the Acre River (left margin), opposite Bom Lugar in concordance with Cabrera (1958).

Geographic distribution
Based on vouchers and literature, T. imperator has been recorded throughout the Brazilian states of Acre and Amazonas limited by the Purus and Acre Rivers (Fig. 10; Appendix 1).In Amazonas, T. imperator is limited by the confluence of the Acre (left bank) and Purus (right bank) Rivers as confirmed by the type series and discussed by Hershkovitz (1979).Records for the species in Acre are based on specimens from Manoel Urbano, Rio Branco, and São Pedro River Basin as observed by Izawa & Bejarano (1981).During analysis of material housed at the MPEG, it was noted that a specimen identified as Midas imperator (MPEG 264) has its provenience tagged as "Cobija Bolivia?".The specimen was identified as S. i. imperator by Hershkovitz (1979) but it undoubtedly has a phenotype Pap.Avulsos Zool., 2023;v.63: e202363005 11/14 of T. subgrisescens.Many primate surveys (Izawa & Bejarano, 1981;Freese et al., 1982;Christen & Geissmann, 1994;Buchanan-Smith et al., 2000) have not reliably recorded the species along the right bank of the Acre River.Indeed, Buchanan-Smith et al. (2000: 366, Table 1) indicated T. imperator in two places in the southern Acre River in Bolivia, but they just stated, "Although locals also reported Saguinus imperator (presumably S. i. imperator) to occur at two sites -Buena Vista and Los Campos -they are considered to be rare, and there has been no previous report of them in the area".Those records must be checked.
All other records of emperor tamarins along the south margin of Rio Acre, including Cobija, clearly present the phenotype of T. i. subgrisescens.The southwestern limit of the species remains unknown.26,1936).A complete synonymy has already been provided by Hershkovitz (1979).

Figure 1 .
Figure 1.Schematic skull of Tamarinus showing the delimitation of the measures used in the morphometric analyses.

Figure 2 .
Figure2.Comparison of some characters between T. i. imperator (A) and T. i. subgrisescens (B) (ventral view).Note the short white hairs tuft, blackish chin, and greyish-brown hairs on the lateral of the arm in A (arrow), and a long tuft of hairs, brown chin, and the lateral fringe of hairs silvery greyish in B (arrow).

Figure 5 .
Figure 5. Rump and dorsal base of the tail grayish in T. i. imperator (A) and rufous in T. i. subgisescens (B).

Figure 6 .
Figure 6.Lateral of the face.Note the hairy ear in T. i. imperator (A) and naked in T. i. subgrisescens (B).

Figure 7 .
Figure 7. Top view of male specimens of the skulls of T. i. imperator (A = MZUSP 11238) and T. i. subgrisescens (B = MZUSP 4806), illustrating the difference in the general aspect.While in T. i. imperator the skull is more delicate and noticeably smaller, the skull of T. i. subgrisescens has both greater size and robustness.

Figure 9 .
Figure 9. Bayesian phylogeny of the genus Saguinus inferred by Cytochrome-b abd 16S data.The posterior probability values are given above the branches.The outgroup was composed of Callimico, Callibella, Callithrix, Cebuella, Mico, and Lentopithecus species.

Table 1 .
Vouchers for each molecular dataset employed, with the respective DNA sequence access numbers in the GenBank.

Table 2 .
Results of Shapiro-Wilk (SW) and t-test (for parametric distribution), and Mann-Whitney (for non-parametric distribution) (P) tests between males and females of T. i. subgrisescens.

Table 4 .
Scores and contributions of each variable in two first principal components (PC) and MANOVA (DF).

Table 6 .
Best partitioning schemes and substitution models.